|
|
|
|
|
|
by sklogic
3846 days ago
|
|
|
Yes, it looks much better than an ad hoc ML. As I said, tricks like Scrap Your Boilerplate are very useful, although they're still not as dense as Nanopass. If a straightforward, depth-first visitor is relatively easy to infer, more complex strategies (e.g., the one you'd need to resolve the lexical scoping) are still a bit messy. And if your visitor is building a new AST out of an old one, things are getting much more tricky. |
|
|